needle-1M-bench (50K): leaderboard
Metric: Overall recall (%). Source: huggingface.co. Saturation forecast: Around December 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 100 |
| 2 | DeepSeek V4 Pro | 100 |
Interactive version: theaggregate.ai/benchmark?slug=needle-1m-bench-50k · How It Works · Data refreshed daily, snapshot 2026-09-26.